Vishal Bhutani

Papers

Total Citations

H-Index

About

Vishal Bhutani is a computer vision researcher whose work centers on advancing unsupervised deep learning for 3D scene understanding from monocular imagery. His most impactful contribution, the 2020 paper "Unsupervised Depth and Confidence Prediction from Monocular Images using Bayesian Inference" (7 citations), introduces a novel framework that integrates Bayesian inference to predict per-pixel depth, confidence maps, and camera pose simultaneously—all without requiring ground-truth depth data. This approach addresses a critical challenge in autonomous navigation and robotics: enabling machines to not only estimate depth from a single camera but also gauge the reliability of those predictions. By outputting confidence alongside depth, Bhutani’s method improves robustness in ambiguous or low-texture regions, a significant step toward safer real-world deployment. His work demonstrates a sophisticated blend of probabilistic modeling and deep learning, offering a principled way to handle uncertainty in geometric estimation. For students and researchers exploring self-supervised depth learning, Bhutani’s framework provides a compelling template for integrating uncertainty quantification into end-to-end vision systems, pushing the boundaries of what can be achieved without expensive labeled data.